Pre-processor definitions slow down the parser significantly

[albertziegenhagel]

Is your feature request related to a problem? Please describe.
I realized that the parser gets slow when one adds a bunch of pre-processor definitions via --pp_defs. As an example I take one of the source files from our code base with ~10'000 lines of code an measure the time of

python -m fortls --debug_filepath=c:\path\to\my\file.F --debug_symbols

which takes ~1 second. But if I add 20 random preprocessor definitions (that are not relevant for this file at all) like

python -m fortls --debug_filepath=c:\path\to\my\file.F --debug_symbols --pp_defs={"DOES_NOT_EXIST1":"1","DOES_NOT_EXIST2":"1","DOES_NOT_EXIST3":"1","DOES_NOT_EXIST4":"1","DOES_NOT_EXIST5":"1","DOES_NOT_EXIST6":"1","DOES_NOT_EXIST7":"1","DOES_NOT_EXIST8":"1","DOES_NOT_EXIST9":"1","DOES_NOT_EXIST10":"1","DOES_NOT_EXIST11":"1","DOES_NOT_EXIST12":"1","DOES_NOT_EXIST13":"1","DOES_NOT_EXIST14":"1","DOES_NOT_EXIST15":"1","DOES_NOT_EXIST16":"1","DOES_NOT_EXIST17":"1","DOES_NOT_EXIST18":"1","DOES_NOT_EXIST19":"1","DOES_NOT_EXIST20":"1"}

this takes ~9.8 seconds.

Describe the solution you'd like
I would expect pre-processor definitions to not have such a severe effect on the parsing speed.